ARTICLE

https://doi.org/10.1038/s41467-019-13490-x OPEN Compact and ultra-efficient broadband plasmonic terahertz field detector

Yannick Salamin 1,5*, Ileana-Cristina Benea-Chelmus 2,4,5*, Yuriy Fedoryshyn1, Wolfgang Heni1, Delwin L. Elder 3, Larry R. Dalton 3, Jérôme Faist 2 & Juerg Leuthold 1*

Terahertz sources and detectors have enabled numerous new applications from medical to communications. Yet, most efficient terahertz detection schemes rely on complex free-space

1234567890():,; optics and typically require high-power lasers as local oscillators. Here, we demonstrate a fiber-coupled, monolithic plasmonic terahertz field detector on a silicon-photonics platform featuring a detection bandwidth of 2.5 THz with a 65 dB dynamical range. The terahertz wave is measured through its nonlinear mixing with an optical probe pulse with an average power of only 63 nW. The high efficiency of the scheme relies on the extreme confinement of the −8 3 terahertz field to a small volume of 10 (λTHz/2) . Additionally, on-chip guided plasmonic probe beams sample the terahertz signal efficiently in this volume. The approach results in an extremely short interaction length of only 5 μm, which eliminates the need for phase matching and shows the highest conversion efficiency per unit length up to date.

1 ETH Zurich, Institute of Electromagnetic Fields (IEF), 8092 Zurich, Switzerland. 2 ETH Zurich, Institute for Quantum Electronics (IQE), 8093 Zurich, Switzerland. 3 Department of Chemistry, University of Washington, Seattle, WA 98195-1700, USA. 4Present address: Harvard John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA 02138, USA. 5These authors contributed equally: Yannick Salamin, Ileana-Cristina Benea-Chelmus. *email: [email protected]; [email protected]; [email protected]

NATURE COMMUNICATIONS | (2019) 10:5550 | https://doi.org/10.1038/s41467-019-13490-x | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13490-x

erahertz (THz)1,2 research is gaining a new momentum substrates are relatively compact and have already demonstrated thanks to recent breakthroughs in the areas of quantum high sensitivities when integrated with plasmonic electrodes19. T 3–5 6 7,8 electrodynamics , ultrafast nanoscopy , spectroscopy , While these detectors feature a broad bandwidth and a very high telecommunications9, and imaging10. Yet, a real outbreak of THz dynamical range at short measurement acquisition times, a fine technology has been hindered so far by the impracticality of balance has to be found between the carrier mobility and lifetime nowadays THz systems that require bulky free-space optics. In to ensure an efficient yet broadband detection20. This compro- addition, the nonlinear mixing with the probe pulse typically mise typically results in an efficiency that drops beyond 1 THz. occurs in exotic material systems, such as e.g. zinc-telluride Conversely, detectors relying on the χ(2) Pockels effect should (ZnTe)11 or gallium arsenide (GaAs)12, and a high sensitivity is offer higher speed. For instance, crystals such as lithium niobate 13 only reached at the expense of a high laser power . Chip-scale, (LiNbO3) or ZnTe have been used to detect THz and mid-IR high-performance and broadband THz systems that exploit well- fields11,21,22. Unfortunately, the χ(2) effect is rather weak and established silicon-on-insulator (SOI) technology and operating comes with a small photon-to-photon up-conversion efficiency, at low optical powers could pave the way to a multitude of new which requires large interaction lengths and in return needs phase applications14. matching of the interacting waves23. Nonetheless, impressive Most widely used room-temperature THz detection approa- results have been demonstrated with field sensitivities equivalent ches rely on photoconduction15,16 or nonlinear parametric up- to a single photon3 or vacuum field fluctuations4. As an alter- conversion in a crystal with χ(2) second-order susceptibility17.In native to inorganic crystals, nonlinear organic (NLO) material both cases, the electric field of a THz wave is measured indirectly, systems demonstrated record high χ(2) nonlinearities22,24. They by means of its modulation onto an optical probe1. Photo- already have demonstrated THz detection up to 15 THz25. Still, conductive antennas (PCA) produced on low-temperature grown electro-optic detection is quite cumbersome as it relies on bulky GaAs (LT-GaAs)12 or indium gallium arsenide (InGaAs)18 free-space optics and rather large nonlinear crystals. Additionally,

a SMF

ETHz,i

THz antenna IIR PPS SMF +Δϕ MZI THz GC IIR+THz

–Δϕ THz GC

b On-chip interferometer c 4LC resonator 50:50

HF antenna PPS

LF antenna

μ 10 μm 50:50 2 m

Fig. 1 On-chip terahertz detector. a The detector consists of a Mach–Zehnder interferometer (MZI) with antenna-coupled plasmonic phase shifters (PPS).

An incident THz field (ETHz,i) introduces, via the linear electro-optic effect, a refractive index change in the gap of the two antenna-coupled phase shifters. This change of refractive index leads to a linear phase delay (±ΔφTHz) of the IR probe pulses (IIR) traveling through the phase shifters. The phase modulation experienced by the IR probe pulses is transformed into an intensity modulation (IIR+THz) at the output of the interferometer. The probe pulses are coupled in and out on-chip silicon (Si) waveguides by means of grating couplers (GC). The organic electro-optic material in the two-phase shifters located in opposite arms of the MZI has opposite polarity to enable push-pull operation. Scale bar is 10 μm. b False color scanning electron image of the fabricated multi-resonant THz detector. The antenna comprises of a high-frequency (HF) antenna and a low-frequency (LF) antenna. Each antenna is resonant within a spectral range. Combined they provide a broadband THz response. c Close-up view of the HF-THz antenna directly coupled to the PPS. Scale bar is 2 μm. SMF single mode fiber, 50:50: multi-mode interferometer.

2 NATURE COMMUNICATIONS | (2019) 10:5550 | https://doi.org/10.1038/s41467-019-13490-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13490-x ARTICLE

a compromise needs to be found between an ideally focused beam The output optical intensity IIR+THz is thus given by and a long interaction length in order to maximize the conversion hiπ fi fi 26 Iin Iin ef ciency. In a rst approach to miniaturize these detectors, an I þ ¼ 1 þ cos þ 2 Δφ  ð1 þ 2 Δn k L Þ antenna was used to collect and confine the THz field to a small IR THz 2 2 THz 2 eff 0 slot volume of 44 μm3 where an electro-optic polymer was located. An ð1Þ electro-optic bandwidth of 1.25 THz was reported with a dyna- = π λ λ = mical power range of ~40 dB. In this way, phase-matching with k0 2 / 0 the wavenumber, 0 1560 nm the probe = μ Δ between the THz wave and optical signal was avoided. Yet, the wavelength, Lslot 5 m the length of the plasmonic slot. neff efficiency was limited by the weak optical probe confinement of is the change of the effective group index of the probe pulse in 35 fi the free-space setup. This led to an inefficient nonlinear inter- each PPS , which changes linearly with the THz eld ETHz,g action with the THz signal and accordingly to a low efficiency. (see the “Methods” section and Supplementary Note 4). A Recently, integrated plasmonic waveguides27 functionalized with detailed analysis of the detector’s performance with respect to nonlinear materials have shown impressive nonlinear χ(2) effi- the MZI operation point is discussed in Supplementary Notes 2 ciencies28. Meanwhile, they have demonstrated good perfor- and 3. mance as electro-optical modulators for high-bit rate systems up The THz antenna consists of two parts, a high-frequency (HF) to 120 GBd29 or for the direct electro-optical up-conversion of 60 and a low-frequency (LF) resonant antenna (see Fig. 1b). The GHz microwave signals30,31. More recently, the frequency combination of the two resonators provides a broadband response of these plasmonic modulators has been shown to frequency response. The HF THz resonator antenna (>1 THz) exceed 500 GHz32. The question that we address here is the fol- contains a four-leaf clover (4LC) structure with the antenna gap lowing one: Can the combination of plasmonics with an organic forming the plasmonic waveguide. The LF resonators are formed nonlinear material offer a simple path towards integrated, effi- by photonic-bandgap (PBG) lines designed to enhance the field at cient broadband THz detectors? lower frequencies (<1 THz) without affecting the HF resonance of To answer this question, we introduce an ultra-compact- the 4LC structure. The exact dimensions of the capacitive and integrated plasmonic THz field detector realized on a silicon inductive elements of the antenna provide a way to tailor the platform. The detector directly measures the electric field frequency response of the detector by its geometry (see amplitude and phase of a THz signal in absolute units (V/m). Supplementary Note 1). Despite the small size of the detector, we show one of the highest A strong THz-to-optical up-conversion efficiency can be conversion efficiencies which, per unit length, is 1000–10,000 expected for multiple reasons in this design. First, the THz times higher than in bulk nonlinear crystals. The high conversion wave collected by the antenna induces a voltage across the efficiency is related to the strong confinement of the THz wave highly sub-wavelength gap and results in a very strong field. and optical probe signal in the nonlinear plasmonic waveguide Further, the field is enhanced by the multi-resonant antenna and a multi-resonant antenna design. Here, multi-resonant design. These two aspects alone provide an enhancement of means that different spectral ranges are covered by individual the THz field in the order of ~1000. Then, the active material in ≈ −1 resonant antennas to provide an overall broadband response. The the slot is highly nonlinear (r33 120 pm V ). Finally, the detector is used to perform THz time-domain electro-optic plasmonic architecture provides an almost perfect overlap of sampling. As a result, we demonstrate an electro-optic bandwidth the probe pulse and the THz wave with the nonlinear material of 2.5 THz with a dynamical range of 65 dB at an optical probe in a slow-light-enhanced plasmonic gap waveguide (see power of 63 nW. Also, we provide a method to retrieve the Supplementary Note 4). The combination of these character- amplitude of an arbitrary input THz pulse. Finally, we compare istics allows a very short active section length, where walk-off the proposed approach with established THz time-domain between the two signals is avoided and the need for phase spectroscopy (TDS) techniques and discuss the advantages of matching is eliminated. In addition, the short plasmonic our implementation. section keeps the optical losses to a miminum. The devices have been realized on a Si wafer using an in- house SOI technology. Details on the fabrication are given Results in the “Methods” section. A close-up view with scanning On-chip THz detection. A schematic view of the on-chip THz electron microscope images of the fabricated MZI and THz detector and its operation principle is given in Fig. 1a. The antenna is shown in Fig. 1b, c. Two THz detectors have been detector comprises a photonic Mach–Zehnder interferometer fabricated with 4LC HF resonances around 1.2 and 2.4 THz, (MZI) in silicon (Si) with THz plasmonic phase-shifters (PPS). respectively. The PPSs consist of a THz antenna with two metallic arms forming a plasmonic slot waveguide filled with a nonlinear χ(2) Time-domain electro-optic sampling experiment. The perfor- 33 fi fi material . An incident THz eld (ETHz,i) excites free charges in mance of the THz eld detector is assessed in an electro-optic the metallic antenna (shown in gold color), which accumulate sampling experiment, where the complex response of our detector across the antenna gap30. The accumulated charges produce an (amplitude and phase) is determined in the time-domain. The fi electric eld ETHz,g across the gap, which changes the refractive measurement setup is depicted in Fig. 2. A femtosecond (fs) laser index in the slot waveguide by means of the linear electro-optic provides fundamental pulses (~150 fs) at an infrared (IR) wave- 34 effect, i.e. the Pockels effect . An optical probe (IIR) launched length of 1560 nm and phase-locked frequency-doubled pulses into the two branches of the MZI is converted into surface (780 nm). The latter pulses are used to generate chopped THz plasmon polaritons (SPPs) at the plasmonic phase shifters where pulses by means of a PCA. The THz pulses are focused by a 90° they experience the refractive index changes. Finally, the SPPs are off-axis parabolic mirror onto the chip. We use as a probe signal converted back to Si waveguides and interfere at the output of the the fundamental pulses. These IR pulses are coupled into a single fi MZI. The intensity at the output of the MZI (IIR+THz) follows a mode ber (SMF) and directly fed to the Si input waveguide of cosine-shaped transfer function34 that depends on the phase the THz detector. At the output, the intensity of the probe signal Δφ delay (± THz) induced by the refractive index change. The MZI is measured by a photodiode. The modulation induced by the is adjusted to operate in the quadrature point, where the output chopped THz signal is detected with a lock-in amplifier as a intensity is linearly related to the incident THz signal amplitude. function of the relative delay between the IR and THz pulses. A

NATURE COMMUNICATIONS | (2019) 10:5550 | https://doi.org/10.1038/s41467-019-13490-x | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13490-x

MZI transfer function 1

Fs laser 12 V 0.5 (1.5 kHz) Lock-in amplifier HWP Reference ΔI 560 nm pp 1 780 nm 0 Transmitted power –1 01 ETHz PCA THz Delay stage detector

ITO SMF PD Fiber coupler

Fig. 2 THz time-domain electro-optic sampling setup. A dual wavelength laser system provides the pump pulses (780 nm) and the optical IR probe signal (1560 nm). The 780 nm signal illuminates the PCA and drives the generation of phase-locked THz pulses. A square bias with a frequency of 1.5 kHz and an amplitude of 12 V is applied to the PCA. The IR probe beam is coupled by means of a single mode fiber (SMF) directly to the on-chip Si input waveguide. A half-wave plate (HWP) controls the linear polarization state of the IR probe to be oriented perpendicular to the plasmonic gap. A motorized delay stage controls the relative delay between the THz signal and IR probe. The probe signal is then guided through the MZI configuration where it probes the amplitude of the THz signal. The output intensity of the probe signal depends on the THz signal as dictated by the MZI transfer function (see inset). The red points depict different relative delays between the optical and THz pulses. Finally, a lock-in amplifier is employed to demodulate the electro-optic signal at the frequency of the driving voltage of the PCA. motorized delay stage controls the relative delay between both output R(ω) and input S(ω) frequency responses and find pulses. ðÞω ðÞ¼ω R : ð Þ Figure 3 presents the main results of the electro-optic- H ðÞω 2 sampling experiment where the experimental complex transfer S function of the detector (Fig. 3d) is measured. Furthermore, we The complex-valued frequency response of the 1.2 THz show that this knowledge can be used to retrieve the time- detector is plotted in Fig. 3d (solid blue lines). The magnitude domain profile of any arbitrary input signal (Fig. 3e). We of the frequency response jjHðÞω reveals two distinct frequency start the discussion with Fig. 3a, which shows the temporal domains (see Fig. 3d) (left y-axis). A first region extends up to fi evolution of an input THz eld (ETHz,i)asdetectedbya about 0.65 THz and obviously shows the response of the LF ZnTe crystal of 200 μmlength(seethe“Methods” section part. A second range with a weaker response extends up to about and Supplementary Note 6). The measured impulse response 1.2 THz. This response can be associated to the 4LC resonator. φ = ω (ETHz,g) obtained with the 1.2 and 2.4 THz detectors are Next, we investigate the phase response H arg[H( )] (see depicted in Fig. 3b. A clear low-noise waveform is found for the Fig. 3d) (right y-axis). The phase is linear and has two distinct two detectors. Each of the two responses exhibit two distinct characteristic slopes in the two frequency domains. This results τ fi pulses spread in time by g.The rst pulse is mainly the from the chirp experienced by the detected pulse, which we impulse response of the HF antenna part optimized for the describe by the group delay reception of signals around 1.2 and 2.4 THz, respectively. This φ pulse is followed by a broader pulse which comes from the LF τ ¼Àd H : ð Þ g 3 antenna response. The spread in time-domain is caused by the dω dispersive nature of the THz antenna. The 2.4 THz detector’s The phase-slopes in Fig. 3d correspond to two distinct group initial pulse has a lower amplitude due to a lower field delays that are offset by 5.25 ps. They are attributed to the amplitude of the source at higher frequencies and due to a impulse response delays of the two antenna parts. We further reduced field enhancement at these frequencies. approximate the measured magnitude frequency response by a The Fourier transform of the recorded time traces R(ω)are step filter combined with an exponential roll-off, see dashed light reported in Fig. 3c. They reveal the frequency responses of blue lines in Fig. 3d. The first step response stems from the LF the measured signals. Both detectors exhibit an electro-optic filter which sets in with a group delay of 5.25 ps. The second step bandwidth exceeding 2 THz with respective dynamic ranges response can be associated to the 1.2 THz 4LC antenna frequency fi 2 of 60 and 65 dB (in eld amplitude squared |ETHz,g| ) measured response. Likewise, we approximate the phase response of the with an optical probe power of 63 nW. The 1.2 THz detector by constant group delays, corresponding to a linear detector features a good signal-to-noise ratio up to 1.2 THz. phase relation (dashed light blue lines). The two slopes of the The 2.4 THz detector drops off beyond 600 GHz. Yet, it phase response correspond to a constant group delay offset of the provides a much better power spectral density between 1.5 and said 5.25 ps. 2.5 THz. The knowledge of the complex-valued frequency response can We derive the detector’s complex transfer function H(ω), with now be used to retrieve the time-domain profile of an arbitrary the purpose to provide a method to remove the dispersion input THz signal from the detected one. When e.g. applying the introduced by the dual-antenna. We start from the measured idealized frequency response H′(ω) to a sampled signal from a

4 NATURE COMMUNICATIONS | (2019) 10:5550 | https://doi.org/10.1038/s41467-019-13490-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13490-x ARTICLE

a s(t) b r(t) 100  1 g 1.2 THz 0.5 ∞ 50 r(t) =∫s(t)h(t – )d –∞ 0 ) 0 –1 –0.5

Amplitude (a.u.) –50

024681012 (kV m 100  Time (ps) g

THz,g 2.4 THz

E 50

0

–50 0 5 10 15 Time (ps)

R() = FT[r(t )] d H()–1.2 THz antenna c R() 0 20 0 Measured Approximated –20 1.2 THz ) 10 R() ) –1 –20 H() = –1 –40 S() 0 –60 0

–40 Phase (rad) –10 –20 2.4 THz Spectrum (dB Hz Spectrum (dB Hz –40 –60 –20 0 0.5 1 1.5 2 –60 Frequency (THz) 0 0.5 1 1.5 2 2.5 3 Frequency (THz)

e s’(t) r’(t) 1 1 FT[r’(t)] s’(t) = FT–1 {} 0.5 0 H’( ) 0 –0.5 Amplitude (a.u.) –1 0 2 4 681012 Amplitude (a.u.) 024681012 Time (ps) Time (ps)

Fig. 3 Experimental results. a Time domain signal of the PCA source measured with a 200 μm ZnTe crystal. b Electric field amplitude of the THz pulse in time domain detected by the MZI detectors with center frequencies of 1.2 and 2.4 THz, respectively. In both cases, the initial pulse of the detected time trace is attributed to the response of the HF-THz antenna component. The following slow THz pulse originates from the response of the LF antenna component. The detected signal is the convolution of the source s(t) with the impulse response of the detector h(t). c The Fourier transform (FT) of the THz time traces for the 1.2 and 2.4 THz detector, respectively. In both cases, the dynamic range is ≥60 dB for an integration time of 20 s per point with an optical probe power of 63 nW. d Measured (solid blue lines) and approximated (dashed light blue lines) complex frequency response of the detector. The frequency response of the detector can be computed from the measured spectrum in c and the spectrum of the known source in a. e Example of a retrieved input signal (black curve) as derived from a measured time trace (green curve) using the approximated frequency response in d.

′ Δ different measurement r (t) obtained with a similar detector, we purpose, we analyze the total peak-to-peak modulation Ipp of can derive the input signal s′(t) (see Fig. 3e). A comparison with the probe intensity (see inset Fig. 2) normalized by the detected the input signal as measured with the ZnTe crystal shows a strong average intensity I. We name this ratio the modulation efficiency η η = Δ similarity. This shows how the approximated transfer function , i.e. Ipp/I. From Fig. 4a one can see that the 2.4 THz would allow to retrieve an unknown THz pulse fairly well. The detector (red square) featuring only 5 μm interaction length is 35 same is valid for the 2.4 THz detector (see Supplementary times more efficient than a 1 mm long bulk ZnTe crystal (blue Note 5). It should be noted that for applications where a strong square). To shed some light on the high efficiency, we plot in group delay should be avoided, a simpler antenna (e.g. bow-tie Fig. 4b the modulation efficiency per unit length as a function of antenna) can be used to minimize the influence of the antenna on the nonlinear detection volume. The efficiency per unit length the measured THz pulse. increases dramatically when transitioning from detection in free space (blue square) to detection in a confined volume (nonlinear Discussion interaction volume) of a few cubic micrometres (green square) fi and then, as in this case, to below 1 μm cube. In fact, in this In Fig. 4, we compare the ef ciency of our detectors with different fi fi fi platforms, namely a 1 mm long bulk ZnTe crystal as a lab stan- con guration the THz eld has been con ned in a compact μ 3 dard and the 3D polymer-antenna26 mentioned before. For this volume of ~0.05 m , which corresponds to a sub-wavelength

NATURE COMMUNICATIONS | (2019) 10:5550 | https://doi.org/10.1038/s41467-019-13490-x | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13490-x

a b 10–2 10–3 75 nm slot 75 nm slot 10–3 10–4

) 2.4 THz MZI antenna –4 100 nm slot –1 1.2 THz MZI antenna 10 m

μ –5

( 10 3D antenna-organic [26]  L ZnTe 1mm @ 780 nm 10–5 2.4 THz MZI antenna per

1.2 THz MZI antenna  10–6 3D antenna-organic [26] 10–6 ZnTe 1mm @ 780 nm 10–7

100 101 102 103 104 10–2 100 102 104 106 108 Interaction length (µm) Nonlinear detection volume (μm3)

Fig. 4 On-chip detector performance. a The modulation efficiency as a function of the nonlinear interaction length is presented for several devices (with a slot width of the PPS of 75 or 100 nm, respectively). For comparison, we show the modulation efficiency of the 3D antenna and a ZnTe crystal of 1 mm length. The rhombus shapes represent extrapolated values from the detection efficiency at 1 mm length. b The modulation efficiency per unit length of interaction distance is represented as a function of the nonlinear detection volume in the different platforms.

−8 λ 3 volume of 10 ( THz/2) . A detailed comparison between dif- Note 2). A detailed analysis of the detector performance with respect to the MZI ferent technologies for THz electric field measurement is pre- operation point is discussed in the Supplementary Note 3. sented in Supplementary Note 7. New opportunities will emerge from the unique advantages Refractive index change. The change of the effective group index of the probe 35 Δ ¼ 1 2 Γ = offered by integrated detection. For instance, more advanced pulse in each PPS is neff 2 nmatr33ETHz;gng c, where nmat 1.77 is the NLO ≈ −1 fi refractive index at optical frequencies, r33 120 pm V is the nonlinear coef cient, detection schemes become possible with the guided optical probe fi Γ ETHz,g is the THz electric eld in the plasmonic slot, and ng c is the product of the by taking advantage of parallelization. A detector array sampling group effective refractive index and the field energy confinement factor (see Sup- the same focused beam could reduce the size mismatch between plementary Note 4). the detector and the focused THz beam spot, including beam steering. In addition, more complex detection schemes are pos- Electro-optic sampling 26 3 . An electro-optic sampling setup as reported in ref. has sible to measure correlation functions . Finally, the advantage of been modified to facilitate the fiber coupling of the probing pulses to and from the low optical powers means a fully integrated THz system on sili- chip. Both the sampling probe pulses and the THz pulses are generated by a fiber- con36 with lasers37–39, frequency combs40 and photodetection41 is coupled IR fs laser at a 90 MHz rate. The large area PCA emitter is biased with a = possible. voltage of Vpp 12 V. The time-domain setup is not purged and the THz pulses are then transmitted through free space for a length of ~50 cm before they reach In conclusion, we have demonstrated a fully monolithic THz the detector. The THz signal was focused on the chip by means of an off-axis field detector that can sample broadband THz pulses with record- parabolic mirror with a focal length of 76.2 mm. The focal length is chosen to allow high efficiency. The compact and integrated architecture is enough space for the fiber coupler and chip mount. The PCA is pumped by 150 fs combined with a single photodiode that is fiber-coupled directly long pump pulses centered around 780 nm. All measurements except the linearity measurements were recorded at a pump power of 76 mW. By sweeping the delay from the chip. This performant yet handy design might enable between the IR probe and the THz pulse, a time-resolved THz field amplitude can new spectroscopic applications, open new avenues for cavity be reconstructed from the different time samples. The time traces for the 1.2 and quantum electro-dynamics in the time-domain4,42 or trigger new 2.4 THz detectors were measured with an integration time of 20 s per point at an venues for THz communications43. Finally, the demonstrated optical probe power of 63 nW (corresponding to 0.7 fJ pulse energy). The recorded time traces were about 15 ps long with time steps of 48 fs. The length of the time platform could pave the way towards a new generation of hand- trace gives a spectral resolution of 62 GHz. Nevertheless, some absorption lines held THz systems operating at lowest power. (1.11, 1.36, 1.6, 2.1, and 2.35 THz) can be identified from the measured spectrum with the 2.4 THz detector and the 1 mm long ZnTe crystal (see Supplementary Note 6). Methods The reference measurements with ZnTe nonlinear crystals with a length of μ Device fabrication. Devices were fabricated in-house on a SOI wafer. First the 200 m and 1 mm were performed in a free-space optical setup including a quarter silicon access waveguides were fabricated by e-beam lithography and dry etching. wave plate, a Wollaston prism and a balanced detector. An optical probe with a An alumina etch stop layer (100 nm) is then deposited. Subsequently, a silicon wavelength of 780 nm was used to maximize phase matching. The optical probe power on the detector was 900 μW and 1.8 mW, respectively. Note that for a dioxide (700 nm)-cladding layer was deposited on top of the silicon waveguide and μ locally etched at the device locations (see Supplementary Note 2). The plasmonic 200 m long ZnTe crystal phase matching is avoided, and we can assume this to be equivalent to the input pulse itself. The measured spectrum with the 1 mm phase shifters and THz antennas were formed by e-beam evaporation of Ti/Au μ (2 nm/150 nm) metal layers and a lift-off process. Finally, the organic electro-optic and 200 m long ZnTe crystals are shown in the Supplementary Note 6. The linearity of our devices was measured by sweeping the optical pump power material composite (3:1 HD-BB-OH:YLD124) was applied by spin-coating on fi the chip and baked. Electric-field poling to induce opposite alignment (push-pull launched on the PCA. The THz eld amplitude depends linearly on the pump configuration) in the two branches of the MZI was carried out by standard power of the PCA. 34 fi ≈ −1 methods . A macroscopic nonlinear coef cient around r33 120 pm V was characterized. Two gap widths are investigated, i.e. 75 and 100 nm. The total Potential of the concept to increase the dynamic range. The dynamical range is optical losses of the integrated plasmonic detector amount to ~31 dB. currently 65 dB. In this implementation it is limited by the optical power level at the output of the chip. However, this can be improved as follows. First, optical losses were rather high in the current chip generation (11.5 dB per grating coupler MZI operation. The THz antenna coupled PPSs on the arm of the MZI introduce a and 8 dB plasmonic losses). Using state-of-the-art silicon foundry coupling Δφ fi 44 phase delay THz proportional to the THz electric eld (ETHz,g). PPSs with schemes would reduce the out-coupling losses by 10 dB . Also, plasmonic losses of opposite crystal orientations in the two branches of the MZI introduce, upon an 3 dB for a 5 μm long device is possible45. Second, higher optical powers would be a incident field, opposite and thus additive phase shifts (push–pull configuration)34. solution to further increase the dynamical range. Yet, in the current imple- – Δφ = Δφ The push pull operation yields a total phase delay of 2 THz. The operating mentation we did not increase the optical probe power in order to stay below the point of the MZI is chosen such that, after interference of the probe pulses at the nonlinear threshold of the silicon waveguide (e.g. two-photon absorption) and the output beam combiner, the output intensity is modulated linearly with the THz damage threshold of the nonlinear material. However, it should be emphasized that electric field (the so-called quadrature point of the MZI) (see Supplementary these are not fundamental limits. The first can be overcome by resorting to a

6 NATURE COMMUNICATIONS | (2019) 10:5550 | https://doi.org/10.1038/s41467-019-13490-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13490-x ARTICLE waveguide material with a larger bandgap (e.g. silicon nitride) and wider cross 17. Wu, Q. & Zhang, X. C. Free‐space electro‐optic sampling of terahertz beams. sections. The latter threshold can be overcome by switching to more recent non- Appl. Phys. Lett. 67, 3523–3525 (1995). linear materials. New organic materials have been shown to work under much 18. Globisch, B. et al. Iron doped InGaAs: competitive THz emitters and detectors higher temperatures46,47. Also, recent barium titanate (BTO) plasmonic mod- fabricated from the same photoconductor. J. Appl. Phys. 121, 053102 (2017). ulators have shown great potential for high-power applications with demonstrated 19. Yardimci, N. T. & Jarrahi, M. High sensitivity terahertz detection through high-temperature stability beyond 250 °C48. With the improvements discussed large-area plasmonic nano-antenna arrays. Sci. Rep. 7, 42667 (2017). above, 15 dB from reduced optical losses and 15 dB from higher optical input 20. Castro-Camus, E., Johnston, M. B. & Lloyd-Hughes, J. Simulation of fluence- powers would increase the output power by up to 30 dB. This should increase the dependent photocurrent in terahertz photoconductive receivers. Semicond. dynamical range from 65 to 95 dB, for shot noise limited detection (see Supple- Sci. Technol. 27, 115011 (2012). mentary Note 8). Finally, one could use a 2 × 2 coupler at the output of the 21. Wu, Q. & Zhang, X. C. Ultrafast electro‐optic field sensors. Appl. Phys. Lett. integrated MZI detector to allow for balanced detection. This would further 68, 1604–1606 (1996). improve the sensitivity of the detector. 22. Han, P. Y., Tani, M., Pan, F. & Zhang, X. C. Use of the organic crystal DAST for terahertz beam applications. Opt. Lett. 25, 675–677 (2000). Data availability 23. Wu, B., Cao, L., Zhang, Z., Fu, Q. & Xiong, Y. Terahertz electro-optic Detailed geometry of the detector is given in Supplementary Fig. 1, and all the sampling in thick ZnTe crystals below the reststrahlen band with a broadband – dimensions are provided in Supplementary Tables 1 and 2. The simulated field femtosecond laser. IEEE Trans. Terahertz Sci. Technol. 8, 305 311 (2018). enhancement of the detectors is given in Supplementary Fig. 2. A detailed MZI transfer 24. Dalton, L. R., Günter, P., Jazbinsek, M., Sullivan, P. A. & Kwon, O.-P. Organic function with indicated operation points is depicted in Supplementary Fig. 3. The MZI Electro-Optics and Photonics: Molecules, Polymers and Crystals (Cambridge tuning structure with measured data is shown in Supplementary Fig. 4. The detector University Press, 2015). efficiency as a function of the MZI operation point is discussed in Supplementary Fig. 5. 25. McLaughlin, C. V. et al. Wideband 15 THz response using organic electro- – The optical mode and THz field confinement in the plasmonic waveguide is shown in optic polymer emitter sensor pairs at telecommunication wavelengths. Appl. Supplementary Fig. 6. The retrieved time-domain input signal amplitude profiles for the Phys. Lett. 92, 151107 (2008). 1.2 and 2.4 THz detectors are shown in Supplementary Fig. 7. The spectral responses of 26. Benea-Chelmus, I.-C. et al. Three-dimensional phase modulator at telecom the 1 mm and 200 μm long ZnTe crystals are given in Supplementary Fig. 8. The electro- wavelength acting as a terahertz detector with an electro-optic bandwidth of – optic sampling characteristics of different platforms are summarized in Supplementary 1.25 terahertz. ACS Photonics 5, 1398 1403 (2018). Table 3. The dynamic range as a function of the probe power and modulation efficiency 27. Gramotnev, D. K. & Bozhevolnyi, S. I. Plasmonics beyond the diffraction – is discussed in Supplementary Fig. 9. The data that support the findings of this study are limit. Nat. Photonics 4,83 91 (2010). available from the corresponding authors upon reasonable request. 28. Ayata, M. et al. High-speed plasmonic modulator in a single metal layer. Science 358, 630–632 (2017). 29. Baeuerle, B. et al. 120 GBd plasmonic Mach–Zehnder modulator with a novel Received: 19 August 2019; Accepted: 4 November 2019; differential electrode design operated at a peak-to-peak drive voltage of 178 mV. Opt. Express 27, 16823–16832 (2019). 30. Salamin, Y. et al. Direct conversion of free space millimeter waves to optical domain by plasmonic modulator antenna. Nano Lett. 15, 8342–8346 (2015). 31. Salamin, Y. et al. Microwave plasmonic mixer in a transparent fibre–wireless link. Nat. Photonics 12, 749–753 (2018). References 32. Burla, M. et al. 500 GHz plasmonic Mach–Zehnder modulator enabling sub- 1. Tonouchi, M. Cutting-edge terahertz technology. Nat. Photonics 1,97–105 THz microwave photonics. APL Photonics 4, 056106 (2019). (2007). 33. Elder, D. L. et al. Effect of rigid bridge-protection units, quadrupolar 2. Mittleman, D. M. Perspective: terahertz science and technology. J. Appl. Phys. interactions, and blending in organic electro-optic chromophores. Chem. 122, 230901 (2017). Mater. 29, 6457–6471 (2017). 3. Benea-Chelmus, I.-C. et al. Subcycle measurement of intensity correlations in 34. Heni, W. et al. Nonlinearities of organic electro-optic materials in nanoscale the terahertz frequency range. Phys. Rev. A 93, 043812 (2016). slots and implications for the optimum modulator design. Opt. Express 25, 4. Benea-Chelmus, I.-C., Settembrini, F. F., Scalari, G. & Faist, J. Electric field 2627–2653 (2017). correlation measurements on the electromagnetic vacuum state. Nature 568, 35. Haffner, C. et al. All-plasmonic Mach–Zehnder modulator enabling optical 202–206 (2019). high-speed communication at the microscale. Nat. Photonics 9, 525–528 5. Riek, C. et al. Direct sampling of electric-field vacuum fluctuations. Science (2015). 350, 420–423 (2015). 36. Harter, T. et al. Silicon–plasmonic integrated circuits for terahertz signal 6. Cocker, T. L., Peller, D., Yu, P., Repp, J. & Huber, R. Tracking the ultrafast generation and coherent detection. Nat. Photonics 12, 625–633 (2018). motion of a single molecule by femtosecond orbital imaging. Nature 539, 37. Guzmán, R., Gordon, C., Orbe, L. & Carpintero, G. 1 GHz InP on-chip 263–267 (2016). monolithic extended cavity colliding-pulse mode-locked laser. Opt. Lett. 42, 7. Eisele, M. et al. Ultrafast multi-terahertz nano-spectroscopy with sub-cycle 2318–2321 (2017). temporal resolution. Nat. Photonics 8, 841–845 (2014). 38. Liu, S. et al. 490 fs pulse generation from passively mode-locked single section 8. Li, X. et al. Terahertz field–induced ferroelectricity in quantum paraelectric quantum dot laser directly grown on on-axis GaP/Si. Electron. Lett. 54, SrTiO3. Science 364, 1079–1082 (2019). 432–433 (2018). 9. Nagatsuma, T., Ducournau, G. & Renaud, C. C. Advances in terahertz 39. Wang, Y. et al. Monolithic quantum-dot distributed feedback laser array on communications accelerated by photonics. Nat. Photonics 10, 371–379 silicon. Optica 5, 528–533 (2018). (2016). 40. Stern, B., Ji, X., Okawachi, Y., Gaeta, A. L. & Lipson, M. Battery-operated 10. Stantchev, R. I. et al. Noninvasive, near-field terahertz imaging of hidden integrated frequency comb generator. Nature 562, 401 (2018). objects using a single-pixel detector. Sci. Adv. 2, e1600190 (2016). 41. Michel, J., Liu, J. & Kimerling, L. C. High-performance Ge-on-Si 11. Winnewisser, C., Jepsen, P. U., Schall, M., Schyja, V. & Helm, H. Electro-optic photodetectors. Nat. Photonics 4, 527–534 (2010). detection of THz radiation in LiTaO3, LiNbO3 and ZnTe. Appl. Phys. Lett. 70, 42. Bayer, A. et al. Terahertz light–matter interaction beyond unity coupling 3069–3071 (1997). strength. Nano Lett. 17, 6340–6344 (2017). 12. Tani, M., Lee, K.-S. & Zhang, X.-C. Detection of terahertz radiation with low- 43. Ma, J. et al. Security and eavesdropping in terahertz wireless links. Nature 563, temperature-grown GaAs-based photoconductive antenna using 1.55 μm 89–93 (2018). probe. Appl. Phys. Lett. 77, 1396–1398 (2000). 44. Vermeulen, D. et al. High-efficiency fiber-to-chip grating couplers realized 13. Yardimci, N. T., Turan, D., Cakmakyapan, S. & Jarrahi, M. A high- using an advanced CMOS-compatible Silicon-On-Insulator platform. Opt. responsivity and broadband photoconductive terahertz detector based on a Express 18, 18278–18283 (2010). plasmonic nanocavity. Appl. Phys. Lett. 113, 251102 (2018). 45. Koch, U. et al. Ultra-compact terabit plasmonic modulator array. J. Light 14. Sengupta, K., Nagatsuma, T. & Mittleman, D. M. Terahertz integrated Technol. 37, 1484–1491 (2019). electronic and hybrid electronic–photonic systems. Nat. Electron. 1, 622–635 46. Deng, G. et al. Monolithic nonlinear optical chromophores with extended (2018). conjugate bridge: large refractive index, high thermal and electro-optic 15. Jepsen, P. U., Jacobsen, R. H. & Keiding, S. R. Generation and detection of stability. Dyes Pigments 164,97–104 (2019). terahertz pulses from biased semiconductor antennas. J. Opt. Soc. Am. B 13, 47. Liu, F. et al. Synthesis of novel nonlinear optical chromophores with enhanced 2424–2436 (1996). electro-optic activity by introducing suitable isolation groups into the donor 16. Jarrahi, M. Advanced photoconductive terahertz optoelectronics based on and bridge. J. Mater. Chem. C 7, 8019–8028 (2019). nano-antennas and nano-plasmonic light concentrators. IEEE Trans. 48. Messner, A. et al. Plasmonic ferroelectric modulators. J. Light Technol. 37, Terahertz Sci. Technol. 5, 391–397 (2015). 281–290 (2019).

NATURE COMMUNICATIONS | (2019) 10:5550 | https://doi.org/10.1038/s41467-019-13490-x | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-13490-x

Acknowledgements Correspondence and requests for materials should be addressed to Y.S., I.-C.B.-C. or J.L. This work was carried out partially at the Binning and Rohrer Nanotechnology Center (BRNC) and in the FIRST lab cleanroom facility at ETH Zurich. The ERC Advanced Peer review information Nature Communications thanks Oleg Mitrofanov, Xiaojun Wu Grants PLASILOR (670478) and MUSiC (340975) are acknowledged for partial funding for their contribution to the peer review of this work. of the work, as well as the US National Science Foundation (DMR-1303080) and the Air Force Office of Scientific Research (FA9550-15-1-0319). Reprints and permission information is available at http://www.nature.com/reprints

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in Author contributions published maps and institutional affiliations. Y.S. and I.-C.B.-C contributed equally to this work. Y.S., I.-C.B.-C., J.F. and J.L. conceived the THz detector concept and designed the experiments. Y.S. designed the THz detectors and performed the optical and THz simulations. I.-C.B.-C. developed the dual frequency Open Access This article is licensed under a Creative Commons electro-optic measurement setup. Y.S. and I.-C.B.-C. implemented the on-chip detection and fiber coupling part of the setup. Y.S. and I.-C.B.-C. performed the measurements. Attribution 4.0 International License, which permits use, sharing, Y.F., Y.S. and W.H. fabricated the samples. W.H. developed the poling procedure for the adaptation, distribution and reproduction in any medium or format, as long as you give activation of the electro-optic organic molecules. D.L.E. and L.R.D. developed the electro- appropriate credit to the original author(s) and the source, provide a link to the Creative optic molecules. Y.S., I.-C.B.-C., J.F. and J.L. analyzed the data. All authors have con- Commons license, and indicate if changes were made. The images or other third party ’ tributed to the writing of the manuscript. material in this article are included in the article s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory Competing interests regulation or exceeds the permitted use, you will need to obtain permission directly from The authors declare no competing interests. the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/. Additional information Supplementary information is available for this paper at https://doi.org/10.1038/s41467- © The Author(s) 2019 019-13490-x.

8 NATURE COMMUNICATIONS | (2019) 10:5550 | https://doi.org/10.1038/s41467-019-13490-x | www.nature.com/naturecommunications